In partnership with
n
Meta just dropped Muse Code a terminal agent that writes, debugs, and ships software while you sleep.
It's not a copilot. It's the whole dev team in a CLI.
And it's already running multi-step tasks across massive codebases.
But wait, there's more:
Jeff Dean left Google after 27 years to automate scientific research itself
A former OpenAI researcher is building tech that turns your thoughts into text
I'm Alex. Welcome to L8R by Innov8.
Let's dive deep 🐰
In today's post:
• Meta ships Muse Spark 1.2 and a terminal agent that codes for you
• Jeff Dean leaves Google after 27 years to automate scientific research
• Ex-OpenAI researcher joins startup building thought-to-text AI
Meta's New Coding AI Wants Your Data for Dirt Cheap
Meta just crashed the AI coding party with a brand new tool.
They launched Muse Spark 1.2 and a terminal agent called Muse Code that builds software for you.
It is a direct attack on tools like Cursor and Claude Code, but with a sneaky twist to grab your coding data.
The details:
On August 5, 2026, Meta dropped Muse Code in beta for macOS and Linux.
The agent runs right in your terminal and uses background helpers that stay active so you do not have to keep guiding them.
It is totally crash-proof. If your system dies, the AI uses a local event log to pick up exactly where it left off.
It comes with smart commands like '/plan' to map out a task and '/grill' to stress-test the code before it runs.
Meta offers a dirt-cheap 'Contributor Tier' at 10 cents per million input tokens, but you must share your prompts and code with them.
Why it matters:
Meta is tired of watching OpenAI and Anthropic rule the coding space.
By offering huge discounts for data sharing, they are solving their biggest problem.
They desperately need high-quality, real-world human code to make their models smarter.
Scraped internet data just does not cut it anymore.
💡 L8R's Take:
Giving Meta all your coding secrets just to save a few bucks is a terrible idea for professional developers.
But hobbyists will absolutely eat up that cheap Contributor tier.
Meta is basically buying the world's best training data for pennies.
From Our Partner
10x the context. Half the time.
Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.
Google's Top Brains Exit to Build Discovery Loop
Google just lost its biggest AI legends.
Jeff Dean and three top researchers are walking out the door to start a new company called Discovery Loop.
This massive brain drain even caused Alphabet's stock to tank.
The details:
Jeff Dean is leaving after 27 years at Google, taking heavy hitters like Sanjay Ghemawat, Oriol Vinyals, and Quoc Le with him.
Their new startup, Discovery Loop, wants to build AI that can run thousands of science experiments all by itself.
Alphabet is actually investing in the startup and giving them computing power for the first year.
The exit triggered a massive shakeup at Google DeepMind, with Demis Hassabis stepping back from daily operations.
Alphabet's stock dropped roughly 4 to 5 percent on August 5, 2026, right after the news broke.
Why it matters:
The smartest minds in AI are tired of big corporate rules.
They want the freedom to build fast without worrying about massive consumer products.
This shows a huge shift where top talent prefers small, focused startups over giant tech monopolies.
💡 L8R's Take:
Google is bleeding its absolute best talent, and that is a terrible sign for Gemini.
You cannot just replace a legend like Jeff Dean.
Discovery Loop will likely build the future of AI research while Google scrambles to pick up the pieces.
From Our Partners
One API. 1,000+ AI Models
Change a single line of code and unlock 1,000+ AI models, smart routing for cost and performance, and persistent agent memory.
Telepathy is the New Typing
🤖 Telepathy is the New Typing
Typing is too slow for the future of AI.
Naomi Bashkansky just ditched her alignment job at OpenAI to build actual telepathy.
She joined a San Francisco startup called Conduit to turn human thoughts directly into text.
The details:
Bashkansky left OpenAI on July 23, 2026, and joined Conduit the very next day.
Conduit makes a headset that reads your brain waves without any surgery or implants.
The team has already collected over 10,000 hours of brain data to train their models.
Their AI can guess what you are thinking with a 45% match right out of the box.
The goal is to let you control computers and AI agents just by thinking about it.
Why it matters:
We are hitting the limits of how fast we can talk or type to machines.
Conduit wants to fix this by plugging AI directly into our brains using huge amounts of data.
If they pull it off, AI stops feeling like a clunky tool. It becomes a natural extension of your own mind.
💡 L8R's Take:
Texting and talking to AI will look like a total joke in ten years.
Conduit is playing the ultimate long game, and I am here for it..
🚀 Quick L8R Summary
Meta's Muse Code: Meta dropped a terminal agent that writes, debugs, and validates code on its own no hand-holding required.
Jeff Dean's exit: Google's legendary chief scientist just left after 27 years to build AI that does research by itself.
Thought-to-text AI: An ex-OpenAI researcher joined Conduit to turn your thoughts into text using brain sensors.
📩 Innathe L8R engane undarunnu 👇?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha L8R il kanaam bie…👋